According to CoinMeta, Bitcoin developers Niklas G and gge stated that recent security incidents such as those involving coldcard, btcpay, server, liquid, and network have shown that the cost of discovering vulnerabilities has significantly decreased. Attackers can now find serious flaws in hardware wallets, Lightning Network, or underlying protocols with relatively low model invocation costs. Currently, Project Loupe, Bitcoin Red Team, and several other developers have used LLM to scan open-source projects like Bitcoin Core, resulting in over 1,000 reports in total. However, most of these reports were false positives or internal API issues, and no high-risk or serious vulnerabilities have been found so far. Gö gge believes that relying solely on LLM scans is not sustainable, as model capabilities continue to improve, costs decrease, and access to such scans is not equal for everyone. A more reliable approach is to strengthen long-term engineering practices such as fuzz testing, property testing, and automated testing.